Issue 1003

[nicholascar]

Adds @base to N3 & Turtle serializations, if base is set in .serialize()

Closes #1003

[coveralls]

Coverage decreased (-0.3%) to 75.623% when pulling a8a0d71 on issue_1003 into acb9d10 on master.

[coveralls]

Coverage increased (+0.008%) to 75.949% when pulling 308abf7 on issue_1003 into acb9d10 on master.

[nicholascar]

close/open Travis

[white-gecko]

I'm not very happy with these changes. For setting the base in the Graph class I would switch the overwriting behavior and for the Turtle serializer I don't like the side effect.

An argument for not changing the Turtle behaviors regarding setting self.base is that we should stick to the same behavior which we have for TriG (

rdflib/rdflib/plugins/serializers/trig.py

Line 48 in 26710b0

def serialize(self, stream, base=None, encoding=None,

) and RDF/XML (

rdflib/rdflib/plugins/serializers/rdfxml.py

Line 48 in 26710b0

def serialize(self, stream, base=None, encoding=None, **args):

)

[white-gecko]

We should also make sure to have the base set for TriX, RDF/XML, and TriG (https://github.com/RDFLib/rdflib/tree/master/rdflib/plugins/serializers)

[nicholascar]

We should also make sure to have the base set for TriX, RDF/XML, and TriG

OK, I'll make the same changes to all

[nicholascar]

I've made the changes I said I would above and got RDF/XML & TriX to work. TriG only 1/2 works since it can't express a base per graph, only one for the whole dataset/conjunctive graph. This means the full solution, which I have no time for, is to reverse QNAME any URI in any individual graph that doesn't use the same base as the dataset/conjunctive graph but the Turtle serializer, which the TriG serializer uses, doesn't keep track of things to make this easy!

So there is a loss of info if multiple graphs within a dataset/conjunctive graph declare base URIs different to the dataset/conjunctive graph and then serialize to TriG... future work!

[white-gecko]

So there is a loss of info if multiple graphs within a dataset/conjunctive graph declare base URIs different to the dataset/conjunctive graph and then serialize to TriG... future work!

I think this is not necessary. The base in general is not relevant info. It is just a convenience for serializations. So I think we do not need to go the extra mile here.

Setting the base on a Graph object is also not necessary but I think a nice feature to give a hint to the Serializers.

An I think the currently set base as attribute on a serialize method should always win over the base set in a graph. serialize(base=base) > graph.base > None.
But the graph forwards is base anyways:
https://github.com/RDFLib/rdflib/blob/master/rdflib/graph.py#L956-L979

[white-gecko]

@nicholascar I have made the changes in the way as I would expect the behavior. Is this ok for you? Maybe @ashleysommer should now review this pullrequest (we should anyhow merge it after the 5.0.0 release).

[nicholascar]

Hi @white-gecko great, I was just about to make the changes you just made! I'll approve and merge as @ashleysommer's not available at the moment.